Plan State Representation Segmentation for Heterogeneous Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current planning techniques face limitations in practicability, scalability, and modularity due to monolithic predicate-based schema in plan state representation, leading to inefficient memory utilization and increased overhead during plan generation, especially when dealing with large and heterogeneous data sources.
Innovation Solution
The system represents a plan state using a combination of logical atoms in a predicate-based schema and database atoms in a non-predicate-based schema, allowing for selective access and caching of database atoms, thereby improving memory utilization and reducing data retrieval overhead.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a monolithic predicate-based schema is used to represent plan state, then the planning system can access information from heterogeneous data sources, but the system suffers from scalability problems due to enormous plan state size and poor memory utilization
Solution Approach 1:
The patent segments the plan state representation into two distinct datasets: a first dataset for logical atoms in predicate-based schema and a second dataset for database atoms in non-predicate-based schema. This segmentation allows the system to handle heterogeneous data sources effectively while improving memory utilization and reducing the overhead of data retrieval, thereby resolving the scalability problem without sacrificing adaptability.
2Ease of operation
If a monolithic predicate-based schema is used for plan state representation, then the planning system can represent world state as a set of predicates, but the system lacks modularity and cannot efficiently handle changes in different data modules
Solution Approach 1:
The patent divides the plan state representation into modular components: logical atoms for dynamic data and database atoms for static data, each with distinct schemas. This modular structure enables independent management and optimization of different data types, making the system more adaptable to changes in specific modules without affecting the entire system.
Solution Approach 2:
The patent extracts static data from the predicate-based schema and places it in a separate non-predicate-based schema for database atoms. This extraction allows the system to maintain the simplicity of predicate-based representation for dynamic data while handling static data more efficiently, reducing overall system complexity.
3Reliability
If complete set of information is loaded into memory during planning, then all data is available for plan generation, but memory utilization becomes inefficient and scalability is reduced
Solution Approach 1:
The patent applies partial action by loading only the necessary subset of information into memory during planning rather than the complete set. The system selectively accesses database atoms only when required, reducing memory consumption while maintaining reliability through on-demand data retrieval.
Solution Approach 2:
The patent extracts the concept of selective data loading from the monolithic approach. By separating static and dynamic data into different schemas, the system can extract and load only the necessary portions of data into memory, improving memory utilization efficiency while maintaining data availability for plan generation.
Data Source
AI summary
A system and method for generating a plan to complete a task by providing a framework facilitating use of heterogeneous data sources without altering a planning algorithm are disclosed. The method includes using a first dataset of logical atoms represented in predicate schema and second dataset of database atoms represented in non-predicate schema, and modifying a grammar rule, a domain definition, and a problem definition, and selecting and executing task methods and task operators to complete a task. Execution of task operator includes verifying a precondition, assigning variables with values when the precondition is valid, and modifying (delete and add) a plan state. Execution of the task method includes verifying a precondition of the task method, assigning variables with values when the precondition is valid, decomposing the task into sub-tasks, assigning arguments of task method to sub-tasks, and adding sub-tasks to a task list. Thereafter, a plan is generated.